International Journal of Engineering & Computer Science IJECS-IJENS Vol:13 No:02 8 134902-8181-IJECS-IJENS © April 2013 IJENS I J E N S An Automated Approach Based On Bee Swarm in Tackling University Examination Timetabling Problem Fong Cheng Weng, Hishammuddin bin Asmuni * Software Engineering Research Group, Software Engineering Department, Universiti Teknologi Malaysia, 81310 UTM Skudai, Johor, Malaysia * Corresponding author. E-mail address: chengweng0410@hotmail.com (CW. Fong), hishamudin@utm.my (H. Asmuni) Abstract -- A recently invented foraging behavior optimization algorithm which is the Artificial Bee Colony (ABC) algorithm has been widely implemented in addressing various types of optimization problems such as job shop scheduling, constraint optimization problems, complex numerical optimization problems, and mathematical function problems. However, the high exploration ability of conventional ABC has caused a slowdown in its convergence speed. Inspired from the Particle Swarm Optimization (PSO) method, an automated approach has been proposed in this study and is named as the Global Best Concept - Artificial Bee Colony (GBABC) algorithm. The algorithm is formulated using the global best concept, which is then implemented into the employed bee phase to incorporate the global best solution information into solutions. This is for the sake of leading the search process towards exploring other potential search regions to locate the best global solution. In addition, to improve its exploitation ability, a local search method has been incorporated into the onlooker bee phase. With the use of the global best concept and local search method, the convergence speed, exploration and exploitation abilities of the basic ABC have been significantly enhanced. Experiments are carried out on standard university examination benchmark problems (Carter’s un-capacitated dataset). Results obtained demonstrate that, generally, the GBABC had outperformed the basic ABC algorithm in almost all instances and its performance is also comparable to other published literature. Index Term-- University examination timetabling, Artificial bee colony algorithm, Hill climbing. 1. INT RODUCT ION Various type of timetabling problems have been addressed by using optimization methods such as job shop scheduling [1-4], flow shop scheduling [5-7], software project scheduling [8], open shop scheduling [9], machine scheduling [10-14], and transportation scheduling [15]. In this paper, the timetabling of university examination is the focus of the study and an overview of related studies can be seen at [16-18]. University examination timetabling is a process of assigning a number of exams into a set of permitted time slots without sacrificing its feasibility; a feasible timetable is one that is clash free. Generally, two distinct types of constraints are encountered in generating a timetable – the hard constraints and soft constraints. Hard constraints must be satisfied under any circumstance in order to preserve the feasibility of the timetable while fulfillment of soft constraints is optional, but its violation should be minimized. This is because a timetable generated is assessed based on its ability to fulfill both hard and soft constraints. Approaches in rectifying university examination timetabling problems vary over a wide rage. From the survey papers [16-18], heuristic approaches that have been applied in solving timetabling problems are mostly based on graph coloring heuristics [16, 19-20]. In recent years, application of meta-heuristic and hybridization approaches have become the main focus and examples of such approaches include the Tabu search [21-25], Simulated Annealing [26-28], Honey Bee Mating optimization [29], Genetic algorithm [30-31], and Great Deluge algorithm [32-36]. Related publications on university timetabling problems can be found in [16-18, 37- 39]. This study, on the other hand, addressed this problem using the Artificial Bee Colony (ABC) algorithm. It is well known that population-based methods like the ABC algorithm must possess adequate exploration and exploitation abilities [40]. The exploration ability allows the bee colony to search and identify possible unknown regions in the search space, whereas the exploitation ability permits the formulation of better solutions based on the information of previous solutions. Ironically, instead of complementing each other, these two abilities are actually in contradiction. Therefore, this study has been conducted to balance these two abilities. The proposed Global Best Concept - Artificial Bee Colony (GBABC) algorithm in this study had been anticipated to improve the convergence speed by enhancing both exploration and exploitation abilities simultaneously with the implementation of the global best concept, which were inspired from the Particle Swarm Optimization (PSO) method and Local Search method. The effectiveness of the proposed algorithm was tested against a set of benchmark datasets - the Carter incapacitated benchmark datasets. Comparison was then made with current state-of-the-art algorithm. In a nutshell, experimental results illustrated that GBABC can generate high quality solutions as compared to basic ABC and the results are also comparable with best reported results. The rest of the paper is organized as follows. Firstly, description on examination timetabling problem is presented